Quickstart
This guide gets you from API key to your first working Milkey MCP workflow in a few focused steps.
Quickstart is the fastest path when you want Milkey inside an MCP-compatible coding agent. You will create a key, add the MCP server, verify the connection, and run a few prompts that prove the setup is working.
Create your Milkey API key
Every Milkey MCP connection starts with a dashboard API key. Create it once, copy it immediately, and keep it somewhere secure.
Key handling
- Create a new key in the dashboard.
- Copy it immediately after creation.
- Store it in a password manager, secret store, or secure local config.
Treat the key like production credentials. Do not commit it to Git or paste it into browser-side code.
Add the Milkey MCP server
Add Milkey as a custom MCP server in your client. Most teams use either a user-wide config for daily work or a project-local config when they want tighter workspace control.
{ "mcpServers": { "milkey": { "url": "https://mcp.vexelityai.com/mcp", "headers": { "Authorization": "Bearer mk_sk_your_api_key_id_your_secret" } } }}Global config
Best when you want Milkey available across multiple repos and day-to-day coding workflows.
Project config
Best when the integration should only apply to one workspace or one team-specific project setup.
https://mcp.vexelityai.com/mcpVerify the setup in your client
Reload the client if needed, then confirm the server is connected before moving on. The goal here is to make sure Milkey is actually available to the agent.
- Open your MCP or integrations panel and confirm that Milkey appears.
- Make sure the server shows as connected rather than unavailable or errored.
- If the connection fails, use the troubleshooting guide.
Use Milkey to explain how the skill system works and tell me which docs page I should read next.
A good result proves the client can see Milkey and resolve a real docs-oriented workflow instead of guessing from model memory.
Use your first Milkey prompts
Once the connection is live, ask for tasks that benefit from current skill guidance. Good first prompts are concrete, technical, and easy to verify.
Use Milkey to find the best skill for React performance optimization in this codebase.
Good for proving skill resolution with a familiar frontend task.
Use Milkey to show me the current FastAPI JWT authentication approach we should follow.
Good for proving that backend setup guidance comes from Milkey rather than stale training data.
Use Milkey to resolve the right skill for database query tuning and summarize the main implementation guidance.
Good for testing a broader architecture or systems workflow.
Next steps
Once quickstart is working, move into the docs that match your workflow and level of control.